UNOMi's AI lip sync reaches Blender and understands Russian

UNOMi has released a version of its AI lip sync for Blender. Until now, the company's automatic speech animation plug-in existed only for Autodesk Maya and Cinema 4D, while support for Blender and Unreal Engine was listed on the developer's website as forthcoming. One of the two promises has now been kept.

The workflow is unchanged: the animator hands the plug-in the audio track of a line along with its text, and the plug-in parses the speech and sets keys on pre-sculpted mouth poses. Manually laying out phonemes on the timeline drops out of the pipeline, after which the artist refines the result themselves. That is the basis of the developer's argument: speed up production while leaving control with the animator.

For Russian-speaking studios, something else matters more. The language list of the Blender version includes Russian: there are eight in total — English, Spanish, Mandarin, Portuguese, French, German, Korean and Russian, with the company promising to add others later. The set differs from what UNOMi previously listed for its 3D plug-in: there it named English, French, German, Italian, Japanese, Spanish and Portuguese, with a note that other languages are supported as well.

What the character needs to provide

The plug-in does not take on modeling: its job is keyframing and animation, and the mouth poses have to be sculpted in advance. For the Maya version, the developer describes the following set of requirements:

• 14 sculpted mouth poses (blendshapes) corresponding to phonetic sounds;
• the audio track of the line and its text transcript;
• a permanent internet connection — it is used to verify the license and to generate the lip sync itself.

Composite characters are supported: eyes, teeth and lips can be split across separate objects. Lip sync for several characters in one scene is also possible, but because of Maya's limitations only one audio track is active during playback. Models are accepted in FBX and GLTF, audio in MP3, WAV and AIFF, and the result is exported back to FBX, GLTF or a QuickTime video. The company has not yet published a separate list of requirements for Blender, so the details of the integration may differ.

The trial version can be downloaded free of charge, while the full version requires a purchase; the product page is UNOMi 3D Lip Sync. In trial mode, the duration of the lip sync is limited to a few seconds. The developer asks Blender users to send in requests for further features.

Why a commercial software vendor is moving into Blender

The plug-in's release looks like part of a broader pivot by the company. UNOMi has stated that within three months its 3D lip sync will turn into UNOMi 3D with a built-in markerless motion capture system working from a webcam or phone; in the plans for that direction, Blender is listed on par with Maya, Cinema 4D and Unreal Engine.

The reason probably lies in the structure of the audience. Small and medium studios, freelancers and indie teams — that is, those for whom automating facial animation saves the most person-hours — increasingly work precisely in Blender. For a supplier of a niche AI tool this is the largest untapped market, and ignoring it any longer was becoming more expensive than porting the plug-in.

Resistance, however, is just as predictable. The Blender community is used to free lip sync add-ons based on phoneme dictionaries and to tools that work offline. A paid license and a mandatory server connection are a familiar model for the Maya market, but for Blender this is a sensitive barrier: without internet and a valid license the plug-in simply will not work. If UNOMi wants to gain a foothold with this audience, the quality of recognition across eight languages will have to be demonstrated convincingly — that is exactly what distinguishes a paid tool from free equivalents.

What matters more to you in lip sync?

• Ready-made keys on mouth poses
• Support for Russian voice-over